�����̨

Thanks to an improved search engine, councillors and policy officers at the Municipality of Utrecht can now find the information they need more quickly. The search algorithm was designed by computer scientist Thomas Schoegje, who earned his PhD on 3 December for his research at the AI Lab for Public Services.

At the Municipality of Utrecht, councillors and policy officers work on various tasks and with different data. This means their information needs are also diverse when, for example, preparing for committee or council meetings. “When employees at the Municipality of Utrecht are looking for information on a specific municipal decision, they can 'Google' it using the search function on the municipal website,” explains Thomas Schoegje. “However, they often receive so many results that it becomes a challenge to determine whether they are relevant to their search.”

Within the AI Lab for Public Services, Schoegje was able to work as a dual PhD candidate at the municipality. This meant he worked both at the municipality and at the university, allowing him to conduct targeted and relevant research. In this role, he developed the search system Ureka, which helps all employees find exactly the right information.

Training AI with domain knowledge

To achieve this, Schoegje 'trained' his search engine using the knowledge of the people who had written the documents. “I modelled the domain knowledge and created an AI system that is able to say: these documents are important for you as a councillor, or as a policy officer. Because the system better shows the context of the information and how the documents are related to each other, it is much clearer than, for example, a system that works based on keywords. A standard solution like that returns many semi-relevant documents.”

In addition, Schoegje developed a historical timeline using AI to analyse, group, and assess the relevance of documents. The timeline allows users to see how the decision has developed over time. “When you click on a search result, you can see where the document fits into the decision-making process. This gives insight into the different steps a decision has gone through, providing more in-depth knowledge and background information.”

Research results in practice

His search engine is now widely used by councillors, Schoegje notes. “I've heard that they use the old system much less now and have largely switched to the new search engine. That's a good sign, as it means my system better meets their needs. For me, that’s the most valuable part of this project—seeing your research results being applied in practice.”

In fact, it’s not only employees at the Municipality of Utrecht who can use the search engine, says Schoegje. “Residents of Utrecht can also use the search engine to create a collection of relevant documents in a timeline. They can only see the public documents, of course.”

Since his promotion, Schoegje has continued to work at the Municipality of Utrecht and is still refining his search engine. “It is now being further developed under the name Raadzaam and is being used by more and more municipalities. Soon, a large municipality will be using it too. In the future, I’d like to expand the search engine with generative AI, so that you could, for example, have a chatbot providing more targeted answers to specific questions.”